Evaluation of QSAR models for predicting mutagenicity: outcome of the Second Ames/QSAR international challenge project

Quantitative structure-activity relationship (QSAR) models are powerful in silico tools for predicting the mutagenicity of unstable compounds, impurities and metabolites that are difficult to examine using the Ames test. Ideally, Ames/QSAR models for regulatory use should demonstrate high sensitivity, low false-negative rate and wide coverage of chemical space. To promote superior model development, the Division of Genetics and Mutagenesis, National Institute of Health Sciences, Japan (DGM/NIHS), conducted the Second Ames/QSAR International Challenge Project (2020-2022) as a successor to the First Project (2014-2017), with 21 teams from 11 countries participating. The DGM/NIHS provided a curated training dataset of approximately 12,000 chemicals and a trial dataset of approximately 1,600 chemicals, and each participating team predicted the Ames mutagenicity of each trial chemical using various Ames/QSAR models. The DGM/NIHS then provided the Ames test results for trial chemicals to assist in model improvement. Although overall model performance on the Second Project was not superior to that on the First, models from the eight teams participating in both projects achieved higher sensitivity than models from teams participating in only the Second Project. Thus, these evaluations have facilitated the development of QSAR models.

Could deep learning in neural networks improve the QSAR models?

Assessing chemical toxicity is a multidisciplinary process, traditionally involving in vivo, in vitro and in silico tests. Currently, toxicological goal is to reduce new tests on chemicals, exploiting all information yet available. Recent advancements in machine learning and deep neural networks allow computers to automatically mine patterns and learn from data. This technology, applied to (Q)SAR model development, leads to discover by learning the structural-chemical-biological relationships and the emergent properties. Starting from Toxception, a deep neural network predicting activity from the chemical graph image, we designed SmilesNet, a recurrent neural network taking SMILES as the only input. We then integrated the two networks into C-Tox network to make the final classification. Results of our networks, trained on a ~20K molecule dataset with Ames test experimental values, match or even outperform the current state of the art. We also extract knowledge from the networks and compare it with the available mutagenic structural alerts. The advantage over traditional QSAR modelling is that our models automatically extract the features without using descriptors. Nevertheless, the model is successful if large numbers of examples are provided and computation is more complex than in classical methods.

A large comparison of integrated SAR/QSAR models of the Ames test for mutagenicity$

Results from the Ames test are the first outcome considered to assess the possible mutagenicity of substances. Many QSAR models and structural alerts are available to predict this endpoint. From a regulatory point of view, the recommendation from international authorities is to consider the predictions of more than one model and to combine results in order to develop conclusions about the mutagenicity risk posed by chemicals. However, the results of those models are often conflicting, and the existing inconsistency in the predictions requires intelligent strategies to integrate them. In our study, we evaluated different strategies for combining results of models for Ames mutagenicity, starting from a set of 10 diverse individual models, each built on a dataset of around 6000 compounds. The novelty of our study is that we collected a much larger set of about 18,000 compounds and used the new data to build a family of integrated models. These integrations used probabilistic approaches, decision theory, machine learning, and voting strategies in the integration scheme. Results are discussed considering balanced or conservative perspectives, regarding the possible uses for different purposes, including screening of large collection of substances for prioritization.

Results of a round-robin exercise on read-across

A round-robin exercise was conducted within the CALEIDOS LIFE project. The participants were invited to assess the hazard posed by a substance, applying in silico methods and read-across approaches. The exercise was based on three endpoints: mutagenicity, bioconcentration factor and fish acute toxicity. Nine chemicals were assigned for each endpoint and the participants were invited to complete a specific questionnaire communicating their conclusions. The interesting aspect of this exercise is the justification behind the answers more than the final prediction in itself. Which tools were used? How did the approach selected affect the final answer?

Integrating QSAR and read-across for environmental assessment

Read-across and QSAR have different traditions and drawbacks. We address here two main questions: (1) How do we solve the issue of the subjectivity in the evaluation of data and results, which may be particularly critical for read-across, but may have a role also for the QSAR assessment? (2) How do we take advantage of the results of both approaches to support each other? The QSAR model starts from the training set. The presence of similar chemicals with property values close to that predicted can support the result. The approach in read-across is the opposite. The assessment is focused on the few substances similar to the target. The data quality of the similar chemicals is fundamental. A risk is poor standardization in the definition of 'similarity', because different approaches may be applied. Inspired by the principles of high transparency and reproducibility, a new program for read-across, called ToxRead, has been developed and made freely available ( www.toxgate.eu ). The output of ToxRead can be compared and integrated with the output of QSAR, within a weight-of-evidence strategy. We discuss the evaluation and integration of ToxRead and QSAR with examples of the assessment of bioconcentration factors of chemicals.

ToxRead: a tool to assist in read across and its use to assess mutagenicity of chemicals

Life sciences, and toxicology in particular, are heavily impacted by the development of methods for data collection and data analysis; they are moving from an analytical approach to a modelling approach. The scarce availability of experimental data is a known bottleneck in assessing the properties of new chemicals. Even when a model is available, the resulting predictions have to be assessed by close scrutiny of the chemicals and the biological properties of the compounds concerned. To avoid unnecessary testing, a read across strategy is often suggested and used. In this paper we discuss how to improve and standardize read across activity using ad hoc visualization and data search methods which use similarity measures and fragment search to organize in a chart a picture of all the relevant information that the expert needs to make an assessment. We show in particular how to apply our system to the case of mutagenicity.

Automatic knowledge extraction from chemical structures: the case of mutagenicity prediction

This work proposes a new structure-activity relationship (SAR) approach to mine molecular fragments that act as structural alerts for biological activity. The entire process is designed to fit with human reasoning, not only to make the predictions more reliable but also to permit clear control by the user in order to meet customized requirements. This approach has been tested on the mutagenicity endpoint, showing marked prediction skills and, more interestingly, bringing to the surface much of the knowledge already collected in the literature as well as new evidence.

CORAL: building up the model for bioconcentration factor and defining it's applicability domain

CORAL (CORrelation And Logic) software can be used to build up the quantitative structure--property/activity relationships (QSPR/QSAR) with optimal descriptors calculated with the simplified molecular input line entry system (SMILES). We used CORAL to evaluate the applicability domain of the QSAR models, taking a model of bioconcentration factor (logBCF) as example. This model's based on a large training set of more than 1000 chemicals. To improve the model is predictivity and reliability on new compounds, we introduced a new function, which uses the Delta(obs) = logBCF(expr)--logBCF(calc) of the predictions on the chemicals in the training set. With this approach, outliers are eliminated from the phase of training. This proved useful and increased the model's predictivity.

Long-term efficacy of infliximab in refractory posterior uveitis of Behcet's disease: a 24-month follow-up study

OBJECTIVES

To evaluate the long-term efficacy and safety of infliximab in patients with Behçet's disease (BD) and refractory bilateral posterior uveitis, and to assess the proportion of relapse-free subjects through months 12 and 24.

METHODS

Open-label, multicentre, 24-month, prospective, follow up study on 12 consecutive patients with BD and refractory posterior uveitis who had failed at least one immunosuppressive drug. At baseline patients received prednisolone 1 mg/Kg/day with rapid tapering and nine infliximab infusions (5 mg/kg) over a 12-month period. Non-responders after the third infusion withdrew from the study. Patients were evaluated for ocular inflammation degree, visual acuity (VA), number of ocular attacks and incidence of adverse events (AEs).

RESULTS

At 12-month visit, 9/12 (75%) patients achieved a complete remission with no relapse during the treatment period. All had a dramatic improvement of ocular inflammation after the first infusion, six were in complete remission after three infusions, and three after four. All these patients suspended corticosteroids at week 22. At 24-month visit, seven out of nine (78%) were still in remission. Mean VA improved from 0.2 +/- 0.6 to 0.5 +/- 0.2 (P < 0.001), and ocular attacks dropped from 40 in the year before therapy to 5 after infliximab cessation (P < 0.001). One patient had a partial remission with two relapses during treatment, and 2/12 (17%) patients showed no improvement. Infliximab was well tolerated with no serious AEs.

CONCLUSIONS

Infliximab is rapidly effective and safe in a high proportion BD patients with refractory posterior uveitis, and may be helpful to prevent recurrences.

Validation of counter propagation neural network models for predictive toxicology according to the OECD principles: a case study

The OECD has proposed five principles for validation of QSAR models used for regulatory purposes. Here we present a case study investigating how these principles can be applied to models based on Kohonen and counter propagation neural networks. The study is based on a counter propagation network model that has been built using toxicity data in fish fathead minnow for 541 compounds. The study demonstrates that most, if not all, of the OECD criteria may be met when modeling using this neural network approach.

Combined administration of alpha-erythropoietin and filgrastim can improve the outcome and cost balance of autologous stem cell transplantation in patients with lymphoproliferative disorders

We compared the use of G-CSF plus EPO in a group of 32 multiple myeloma and lymphoma patients with historical controls receiving G-CSF alone. Haemopoietic reconstitution was significantly faster in patients receiving G-CSF+EPO (group B), with a median time of 10 days to achieve an ANC count >0.5 x 10(9)/l, compared to 11 days in the historical group (A). The median duration of severe neutropenia (ANC count <100/ml) was significantly shorter in group B compared to group A; platelet counts >20 x 10(9) and >50 x 10(9)/l were achieved at days + 13 and + 17, respectively in group B, compared to days + 14 and + 24, respectively, in group A (P = 0.015, 0.002) patients. The transfusion requirement was reduced in group B, with 0 (0-6) RBC units and 1 (0-5) platelet unit transfused in group B vs 2 RBC (0-9) and 2 platelet units (0-8) in group A. Median days of fever, antibiotic therapy and hospital stay were reduced in group B (9.5 days vs 22). The mean cost of autotransplantation per group A patient was 23,988 Euro, compared with 18,394 Euro for a group B patient. Our study suggests that the EPO + G-CSF combination not only accelerates engraftment kinetics, but can also improve the clinical course of ASCT.

Predicting logP of pesticides using different software

We compared experimental and calculated logP values using a data set of 235 pesticides and experimental values from four different sources: The Pesticide Manual, Hansch Manual, ANPA and KowWin databases. LogP were calculated with four softwares: HyperChem, Pallas, KowWin and TOPKAT. Crossed comparison of the experimental and calculated values proved useful, especially for pesticides. These are harder to study than simpler organic compounds. Structurally they are complex, heterogeneous and similar to drugs from a chemical point of view. They offer an interesting way to verify the goodness of the different methods. Other studies compared several logP predictors using a single set of experimental values taken as a reference. Here we discuss the utility of the different logP predictors, with reference to experimental data found in different databases. This offers three advantages: (1) it avoids bias due to the assumption that one single data set is correct; (2) a given predictor can be developed on the same data set used for evaluation; (3) it takes account of experimental variability and can compare it with the predictor's variability. In our study Pallas and KowWin gave the best results for prediction, followed by TOPKAT.

Tuning neural and fuzzy-neural networks for toxicity modeling

The need for general reliable models for predicting toxicity has led to the use of artificial intelligence. We applied neural and fuzzy-neural networks with the QSAR approach. We underline how the networks have to be tuned on the data sets generally involved in modeling toxicity. This study was conducted on 562 organic compounds in order to establish models for predictive the acute toxicity in fish.

Incidence and clinical characteristics of posttransplant lymphoproliferative disorders: report from a single center

In the period 1973-1998, among 2139 allograft recipients treated with standard immunosuppression, posttransplant lymphoproliferative disorders (PTLD) developed in 19 patients (0.9%): one plasmacytic hyperplasia, two polymorphic PTLD, one myeloma, and 15 lymphomas. PTLD developed 1 year after transplantation (tx) in 14 patients. Five patients were diagnosed at autopsy, 2 were lost to follow up, 3 died before therapy could be instituted, and 1 patient has just started chemotherapy. Of the 8 evaluable patients, 2 received acyclovir and are alive in complete remission (CR) and 6 received chemotherapy +/- surgery. Of these 6, 4 died of lymphoma and/or infection, 1 died of unrelated causes in CR, and 1 is alive in CR. PTLD is a severe complication of tx, usually running an aggressive course which may preclude prompt diagnosis and treatment. Nevertheless, therapy is feasible and must be tailored on the histologic subtype. Seventy-four percent of patients were diagnosed with late-onset PTLD stressing the need for long-term follow up.

Predictive carcinogenicity: a model for aromatic compounds, with nitrogen-containing substituents, based on molecular descriptors using an artificial neural network

A back-propagation neural network to predict the carcinogenicity of aromatic nitrogen compounds was developed. The inputs were molecular descriptors of different types: electrostatic, topological, quantum-chemical, physicochemical, etc. For the output the index TD50 as introduced by Gold and colleagues was used, giving a continuous numerical parameter expressing carcinogenicity. From the tens of descriptors calculated, principal component analysis enabled us to restrict the number of parameters to be used for the artificial neural network (ANN). We used 104 molecules for the study. An Rcv2 = 0.69 was obtained. After removal of 12 outliers, a new ANN gave an Rcv2 of 0.82.

Computational predictive programs (expert systems) in toxicology

The increasing number of pollutants in the environment raises the problem of the toxicological risk evaluation of these chemicals. Several so called expert systems (ES) have been claimed to be able to predict toxicity of certain chemical structures. Different approaches are currently used for these ES, based on explicit rules derived from the knowledge of human experts that compiled lists of toxic moieties for instance in the case of programs called HazardExpert and DEREK or relying on statistical approaches, as in the CASE and TOPKAT programs. Here we describe and compare these and other intelligent computer programs because of their utility in obtaining at least a first rough indication of the potential toxic activity of chemicals.

[Randomized study of the antiemetic efficacy of alizapride versus alizapride + dexamethasone]

We made a study on patients suffering from neoplasias treated with cisplatinum to compare antiemetic efficacy of alizapride or associated to dexamethasone. We divided patients in two groups. We somministred alizapride alone to the first group and alizapride plus dexamethasone to the second one. We evaluated presence or absence of emesis, its intensity and eventual tossic effects due to the drugs through an autocompilative questionnaire given to the patients. Our results didn't show important statistical differences between the two groups, though association with dexamethasone gets better emesis control. Alizapride anyway, didn't present important collateral effects, particularly extrapyramidal ones.

A serial model for computer assisted medical diagnosis

In this paper we discuss a simple model for an interactive consultation system for medical applications. That system was developed to provide consultative advice on diagnosis of arthritis, but it could be easily applied to a different class of diseases. We examine in particular how the determination of the appropriate sequence of diagnostic tests that have to be performed on the patient can be optimised by the use of a decision table, in which the medical knowledge is shown. We begin with a review of features which were seen to be essential to a consultation system for medical applications and we suggest how these requirements imply program design criteria. This is followed by a formal explanation of the proposed model and its fundamental assumptions. The last part of the paper is then devoted to a report of our experience with the experimental computer system.